Data quality assessment framework to assess electronic medical record data for use in research

Area of Science:

• Health Informatics
• Data Science
• Clinical Research

Background:

• Electronic medical records (EMR) offer vast clinical data for patient outcomes and health systems research.
• Assessing the suitability of aggregated EMR data for research is crucial.
• Existing data often requires rigorous quality evaluation before use.

Purpose of the Study:

• To present a stepwise framework for initial data quality assessment of multiple data sources.
• To demonstrate the framework's application on a longitudinal electronic health record repository.
• To evaluate the suitability of electronic health record data for research purposes.

Main Methods:

• A six-step data quality assessment framework was developed: preliminary analysis, documentation-longitudinal concordance, breadth, data element presence, density, and prediction.
• The framework was applied to the Transport Data Mart, a repository of over 28,000 interhospital transfer patient records.
• The analysis included EMR data from sending, transport, and receiving hospitalizations.

Main Results:

• A 93% automatic matching rate was achieved for hospital encounters, with 2832 records mapped to both sending and receiving hospitals.
• Limited data drop-off was observed when increasing matching criteria, indicating consistent data completeness in a proportion of records.
• Measures of central tendency showed right skewness, suggesting a small subset of records contained the most repeated measures.

Conclusions:

• The proposed six-step framework is effective for establishing reproducible metadata for longitudinal data repositories.
• Practical challenges in data quality assessment require addressing, particularly establishing metrics for EMR data inclusiveness.
• Further research is needed to define benchmarking standards for acceptable EMR data quality in research applications.